Portfolio

portfolio.

Evidence of judgment, not a list of builds.

AI platform judgment, AI-assisted expense capture and a live group ledger. Three ways I turn messy workflows into products people can understand and trust.

Projects

Product blueprint · Agentic AI

From Prototype to Production

A product blueprint for making enterprise AI agents trustworthy, observable, reusable and cost-aware.

Problem
Agent demos stall before production.
Decision
Build the platform around the agent.
Signal
Grounded in owning an enterprise GenAI platform roadmap.
Agentic AIMCPLLM EvaluationPlatform Product
NotSplitwise on a phone, three screens: capturing the note “1240 lunch, Asha paid, me + Ben”, a review screen asking who paid, and an export reference for manual entry
Built & reviewed
2026
AI PRODUCT · BUILT & REVIEWED

NotSplitwise

Capture trip expenses quickly. Review what AI understood before exporting a clean reference.

Problem
Expense notes mix amounts, payers and people.
Decision
AI suggests. The person confirms.
Signal
Working capture, review and export; independently tested recovery paths.
AI ExtractionHuman ReviewNext.js
Pokerbook active session “Rana” with five players, their buy-in counts, a ₹16,000 table bank, and copy-link and add-player controls
Live Product
2026
Fintech & Systems · Consumer UX · Shipped Product

Pokerbook

A live group ledger for poker buy-ins, cash-outs and settlements.

Problem
Home-game settlements get messy.
Decision
Make mismatches explicit before settling.
Signal
Live, used by 50+ people (self-reported).
Live LedgerSettlement LogicNext.js
Get in touch

Got something messy?

I own an enterprise GenAI platform roadmap and still build my own products. Open to remote roles and to relocating to Europe, the UAE or Singapore.